Back in April, I pinged this list[1] for help in specifying a ZFS server that would handle high-capacity reads and writes. That server was finally built and delivered, and I''ve blogged the results[2] as part of a larger series[3] about that server. [1] http://www.opensolaris.org/jive/thread.jspa?threadID=28720 [2] http://lindsay.at/blog/archive/2007/09/14/pause-for-testing.html [3] http://lindsay.at/blog/archive/tag/zfs-performance/ Initial testing has raised some questions against my initial understanding: - With RAID-Z groups, why would I be seeing sustained write performance consistently at 75% of the speed of sustained reads? - With Random I/O, I see performance increasing logarithmically with the number of disks (or vdevs), whereas I was expecting something a bit more linear. Has anyone else run into this? If it''s not mechanical (vibrations due to 15 or so disks bouncing all over the place), what else could it be? Many thanks, adam